Knowing the structure of proteins is key to figuring out their functions and the mechanisms by which they operate. This information is essential for drug target selection and drug design, as well as for fundamental understanding of both disease processes and the normal operation of cells. Unfortunately, experimental methods for determining protein structure cannot keep up with the rapid growth in protein sequence data, so computational methods are needed to predict the structure from the sequence data. The goals of this project are to produce the world's best program for automatic prediction of protein structure from sequence and to make the tool available to biologists, both on the web and as a distributable software package. The project combines three different, but complementary, approaches to protein-structure prediction: 1D prediction of local structural properties using neural nets, fold-recognition using hidden Markov models, and conformation generation and scoring using fragment packing and an emprical energy function.
Showing the most recent 10 out of 13 publications
